Information Theory & Network Economics
Information Rules: A Strategic Guide to the Network Economy
Shapiro and Varian's point for investors: information goods cost a fortune to make once and almost nothing to copy, so the cost structure, not the product, decides who earns the margin. That structure can be measured.
The big picture
The authors argue that the internet did not repeal economics; it changed which economics bind. Software, data and media carry a large first-copy cost (the fixed spend to produce the first unit) and a marginal cost (the cost of one more unit) close to zero. With that shape, competition pushes price toward marginal cost, so sellers who survive do it by pricing differently to different buyers, by making it costly for customers to leave, or by owning the standard everyone else has to plug into. The book's core bet: in network markets, strategy is mostly about cost structure, switching costs and standards, and those three can be analyzed with ordinary tools.
Why it matters now: large AI models are the purest information good yet built. Training is the first-copy cost, inference is the marginal cost, and the price per token for a given level of model capability has fallen steeply since 2023. The book's framework is a way to read which firms in that chain carry operating leverage (profits that swing more than revenue) and which ones are being competed down toward cost.
The 3 strategic pillars
-
Cost structure is the product
When the first copy is expensive and every further copy is nearly free, average cost keeps falling with volume and scale becomes the main advantage.
Measure it with operating leverage: the share of costs that stay fixed as revenue moves. A high contribution margin (revenue left after variable costs) over a thin operating margin means each extra dollar of sales drops mostly to profit, and each lost dollar comes mostly out of it.
-
Price by value, not by cost
Because cost gives no floor worth anchoring to, sellers set price by what each buyer group is willing to pay, and offer versions to let buyers sort themselves.
Versioning means shipping the same core good in tiers (speed, limits, features) so a price-sensitive buyer and a demanding buyer each pay close to their own value. The test is simple arithmetic: does the tiered menu collect more revenue than the best single price?
-
Lock-in and positive feedback
Switching costs tie customers to a supplier, and network effects make a product more valuable the more people use it, so markets tend to tip toward one or two winners.
The total cost of a switch is what the customer pays to move plus what a rival pays to win them. A customer is worth, at most, the margin you earn from them plus that total switching cost; bidding more than that for them is how standards wars destroy value.
What a Closelooknet reader does with it
The working use is a cost-structure read before you form a view on any software, platform or AI name: split the income statement into fixed and variable costs, compute contribution margin, operating leverage and break-even, then ask how far revenue can fall before profit disappears. The mistake this prevents is treating a high-margin information business as stable when its margin is mostly leverage, which cuts both ways. The versioning and switching-cost sheets add the second question: is the pricing power real, or is it a lock-in that a cheaper rival or an open standard can dissolve?
The bridge to the Closelooknet approach
Closelooknet reads the AI build-out through the same lens. Inference Economics is the book's cost split restated for AI: training behaves like the first-copy cost, inference like the marginal cost that scales with every query. SaaSpocalypse is a switching-cost argument: if agents lower what it costs a customer to leave, the subscription lock-in behind SaaS pricing weakens. Cloudflare as Agentic Toll Booth is the positive-feedback case, a platform that gains value as more traffic runs through it. For the information-theory side of the same cluster, see The Information; for the durability question, the glossary entries on moats and net retention give the measurable versions.
Action-Kit — from theory to practice
Tooling & data
| What you need | Where to get it | Cost |
|---|---|---|
| Income statements with cost lines (revenue, cost of revenue, R&D, S&M, G&A) Raw inputs for the fixed-versus-variable split and the operating-leverage calculation | SEC EDGAR full-text filings (10-K / 10-Q) The split into fixed and variable is your judgment call; the pack keeps it as an explicit input so the assumption stays visible. | Free |
| Standardized multi-year financials Quick history of revenue and operating income to estimate operating leverage from actual year-on-year changes | stockanalysis.com Free tier shows several years of annual statements; longer history and export need the paid tier. | Freemium |
| Model token price lists Reference points for how the marginal cost of AI output is priced and tiered (a live example of versioning) | Published API pricing pages of the model providers, e.g. Anthropic and OpenAI Record the date you read a price list; token prices change often. | Free |
The formulas
-
Degree of operating leverage (DOL)
DOL = %ΔEBIT / %ΔRevenue ≈ Contribution margin / Operating margin- Contribution margin — (Revenue − Variable costs) / Revenue
- Operating margin — EBIT / Revenue
- Or two periods of revenue and EBIT for the observed version
A DOL of 4 means a 10% revenue change moves operating profit by about 40%, in either direction. The ratio explodes near break-even, which is the point.
-
Break-even volume
Break-even units = Fixed cost / (Price − Marginal cost)- Fixed cost — first-copy and other period costs
- Price — revenue per unit (seat, subscription, million tokens)
- Marginal cost — cost to serve one more unit
In revenue terms: break-even revenue = Fixed cost / Contribution margin.
-
Customer value with switching cost
Max value of a customer ≈ PV(margin per period) + total switching cost; total switching cost = customer's cost to move + rival's cost to win- Margin per period and expected retention
- Discount rate
- Customer-side switching cost (migration, retraining, data)
- Supplier-side cost to acquire (discounts, incentives, integration)
The pack computes a simple lifetime value and shows how much of it rests on lock-in rather than on product margin.
-
Versioning revenue check
Uplift = Σ tier revenue (each segment picks its best tier) − best single-price revenue- Segment sizes
- Willingness to pay per segment for each tier
A positive uplift only holds if the lower tier does not pull high-value buyers down; the sheet flags that case.
Applied Pack · free members
Varian Applied Pack
The network-economy arithmetic as a working calculator: split your own cost lines into fixed and variable, read operating leverage and break-even, test a tiered price menu and put a number on lock-in.
- Varian_Operating_Leverage_Calculator.xlsx — cost structure, contribution margin, DOL, break-even and a revenue-shock sensitivity table, plus a three-tier versioning sheet and a switching-cost / lifetime-value sheet, all with live formulas and amber input cells
- varian_screener.py — stdlib-only screener: feed it a CSV of your own income-statement lines, get contribution margin, DOL, break-even revenue and a ranked table
- income_lines_sample.csv — example input with EXAMPLE_ rows showing the expected columns
- README.txt — column reference, how to run, and the educational-use disclaimer
Pack security
Macro-free Excel · plain-text Python you can read before you run it · no installers, no network access — the code works only on files you provide. Served only from closelook.net; we never distribute through download portals or email attachments. How to verify in 30 seconds →
SHA-256 c692943de16563c8680d788de83ed4b1ef92fdac0d97608771d4af5dc68968bb
The pack is a free-membership download — no card, free forever. Membership also delivers the Daily Pulse and Weekly Signal to your inbox.
Join the Look — free → Already a member? Sign in →
Educational templates — a research diary companion, not investment advice.
Closelooknet publishes a market diary, not investment advice. This condensed read restates the book's ideas in our own words for education — for the author's full argument, go to the source.